A Data Processing Method and System for Leakage Fault Source Tracing Based on Time-Sequence Correlation

By passively collecting leakage current time-series data in the power distribution network, establishing electrical topology constraints and performing feature decoupling, calculating directional information flow and correcting response delay, and combining zero-sequence conservation residuals and operational transient shielding results, the accuracy and efficiency of leakage fault tracing are solved, and the accurate location of the main anomaly source is achieved.

CN123066482APending Publication Date: 2026-11-10SHANGHAI YIDIAN XINSEN TECH DEV CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611255597.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-18
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing leakage current monitoring methods are difficult to accurately determine the starting location of leakage current anomalies and are easily affected by load switching, switching operations and power frequency disturbances of nonlinear loads, making it difficult to trace the source of leakage current faults.

Method used

By passively collecting leakage current time-series data from multiple monitoring points, electrical topology constraints are established, dual-branch leakage characteristics are decoupled, directional information flow and correction response delay are calculated, and combined with the zero-sequence conservation residuals of parent and child loops and the results of operational transient masking, a comprehensive score of the main anomaly source is generated, and the main anomaly source monitoring point and anomaly response time-series chain are identified.

Benefits of technology

Without injecting test signals into the power distribution line or controlling the opening and closing of switches, the main abnormal source monitoring point can be accurately identified, reducing the impact on the original power supply status and improving the accuracy and efficiency of tracing leakage faults.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123066482A_ABST
    Figure CN123066482A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of data processing technology, specifically disclosing a data processing method and system for tracing leakage current faults based on time-series correlation. The method includes: passively collecting leakage current time-series data from multiple monitoring points under the conditions that the power distribution network maintains its original operating state and no test signals are injected into the power distribution lines or the switches are not controlled to open or close, and establishing electrical topology constraint relationships between the multiple monitoring points. The technical problem to be solved by this invention is how to accurately identify the main anomaly source monitoring point and generate an anomaly response time-series chain based on the leakage current time-series data passively collected from multiple monitoring points, combined with electrical topology constraints, measurement link response delay correction, parent-child loop zero-sequence conservation residuals, and operational transient shielding results, without injecting test signals into the power distribution lines or controlling the switches to open or close.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for tracing leakage faults based on time-series correlation. Background Technology

[0002] In scenarios such as power distribution rooms, building distribution boxes, industrial low-voltage distribution cabinets, and temporary construction power systems, multiple branch circuits are usually connected by the same upstream circuit, the same busbar, or the same grounding system. When any circuit experiences a leakage current anomaly, the leakage current sampling values ​​of the upstream circuit, the same-level branch circuit, or related circuits along the common grounding path may change synchronously or with a lag. Existing leakage current monitoring methods mostly use the leakage current threshold of a single monitoring point as the alarm basis. This can determine whether a monitoring point exceeds the limit, but it is difficult to determine whether the monitoring point is the actual starting point of the leakage current anomaly or a response point formed by the influence of other circuits. For multi-monitoring point leakage current monitoring systems, different leakage current sensors, signal conditioning circuits, and data acquisition channels have different response delays. Directly comparing the anomaly occurrence time of each monitoring point may lead to errors in judging the order of anomaly responses. Meanwhile, non-fault factors such as load switching, switch operations, and power frequency disturbances from nonlinear loads also exist in the power distribution network. These factors may cause a short-term increase in leakage current at some monitoring points. If judgment is based solely on time-series correlation or current change amplitude, short-term operational disturbances can easily be identified as leakage anomalies. Furthermore, the statistical correlation between multiple monitoring points may not necessarily conform to the actual power supply connection relationship. Without constraints such as parent-child circuit relationships, same-level branch relationships on the same busbar, and common grounding path relationships, electrically invalid abnormal response relationships can easily be generated. Therefore, without injecting test signals into the distribution lines, controlling the switches to open or close, or changing the original operating state of the power distribution network, how to distinguish the main anomaly source monitoring point from the affected response points based on passively collected multi-monitoring point leakage current time-series data, and form an abnormal response time-series chain that conforms to electrical connection relationships and response sequence relationships, is a problem that needs to be solved in tracing leakage faults in low-voltage power distribution networks. Summary of the Invention

[0003] This invention provides a data processing method and system for tracing leakage current faults based on time-series correlation. The technical problem to be solved by this invention is how to accurately identify the main abnormal source monitoring point and generate an abnormal response time sequence chain based on the leakage current time-series data passively collected from multiple monitoring points, combined with electrical topology constraints, measurement link response delay correction, parent-child loop zero-sequence conservation residuals, and operational transient shielding results, without injecting test signals into the power distribution line or controlling the switch to perform opening or closing.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A data processing method and system for tracing leakage current faults based on time-series correlation includes: passively collecting leakage current time-series data from multiple monitoring points under the condition that the power distribution network maintains its original operating state and no test signals are injected into the power distribution lines, and no switches are controlled to perform opening or closing operations, and establishing electrical topology constraint relationships between the multiple monitoring points; decoupling the leakage current time-series data by dual-branch leakage characteristics to obtain trend characteristics, power frequency characteristics, and branch mismatch indices for each monitoring point; calculating the directional information flow and corrected response delay between each monitoring point based on the trend characteristics, and generating electrical topology constraint candidate edges in combination with the electrical topology constraint relationships; calculating the comprehensive score of the main anomaly source for each monitoring point based on the electrical topology constraint candidate edges, the branch mismatch indices, the zero-sequence conservation residuals of parent and child circuits, and the operational transient shielding results, and determining the main anomaly source monitoring point and the anomaly response time-series chain based on the comprehensive score of the main anomaly source.

[0005] In one aspect of the present invention, establishing the electrical topology constraint relationship between the plurality of monitoring points includes: determining the parent-child loop relationship, the same-level branch relationship under the same bus, and the common grounding path relationship between each monitoring point according to the power supply connection relationship of the power distribution network, and generating an electrical topology adjacency matrix based on the parent-child loop relationship, the same-level branch relationship under the same bus, and the common grounding path relationship; wherein, generating an electrical topology constraint candidate edge between two monitoring points is allowed only when there is an electrical connection mark in the electrical topology adjacency matrix that allows edge establishment.

[0006] In one aspect of the present invention, passively acquiring leakage current timing data from multiple monitoring points includes: acquiring zero-sequence leakage current signals of corresponding circuits through leakage current sensors arranged at each monitoring point; inputting the zero-sequence leakage current signals into a multi-channel synchronous sampling analog-to-digital converter after signal conditioning to obtain digital leakage current sampling values ​​with a unified timestamp; and forming a similar leakage current timing group by combining the digital leakage current sampling values ​​obtained from each monitoring point under the same sampling clock; wherein, the similar leakage current timing group does not include impedance spectrum data obtained by actively injected test signals, nor does it include state transition data obtained by switch opening or closing disturbances.

[0007] In one aspect of the invention, the dual-branch leakage current feature decoupling of the leakage current time series data includes: in the trend feature branch, performing sliding effective value calculation and trend filtering on the leakage current time series data to obtain trend sampling values, and calculating the time series change rate and time series cumulative value based on the trend sampling values; in the power frequency feature branch, performing spectrum analysis on the leakage current time series data to extract the fundamental leakage current amplitude and the second harmonic leakage current amplitude; and generating a branch mismatch index based on the time series change rate, the fundamental leakage current amplitude, and the second harmonic leakage current amplitude, wherein the branch mismatch index is used to characterize the prominence of trend anomalies relative to power frequency periodic disturbances.

[0008] In one aspect of the present invention, calculating the directional information flow and corrected response delay between monitoring points based on the trend characteristics includes: constructing a state space reconstruction vector based on the trend sampling values ​​of each monitoring point, and calculating the normalized value of the transfer entropy between any two monitoring points based on the state space reconstruction vector, and using the normalized value of the transfer entropy as the directional information flow value; determining the observation response delay based on the time delay correlation result between the trend sampling values ​​of the two monitoring points; correcting the observation response delay using the measurement link response delay of the leakage current sensor, signal conditioning circuit and data acquisition channel corresponding to each of the two monitoring points, to obtain the corrected response delay; wherein, only when the directional information flow value exceeds the directional information flow edge establishment threshold and the corrected response delay is positive, the corresponding two monitoring points have the timing conditions for generating candidate abnormal response edges.

[0009] In one aspect of the invention, generating candidate edges for electrical topology constraints in conjunction with the electrical topology constraint relationships includes: matching candidate abnormal response edges that satisfy the timing conditions with the electrical topology adjacency matrix; if there is an electrical connection marker between the starting and ending monitoring points of the candidate abnormal response edge that allows edge establishment, then the candidate abnormal response edge is retained as a candidate edge for electrical topology constraints; if there is no electrical connection marker that allows edge establishment, then the candidate abnormal response edge is deleted; thereby filtering out pseudo-abnormal response edges that only have statistical correlation but do not satisfy the electrical connection relationships of the distribution network.

[0010] In one aspect of the invention, the generation method of the zero-order conservation residual of the parent-child loop includes: for monitoring points with a parent-child loop relationship, calculating the zero-order trend consistency deviation between the parent monitoring point and the child monitoring point set based on the trend sampling value of the parent monitoring point, the trend sampling value of each child monitoring point in the child monitoring point set, and the proportional correction coefficient between the parent monitoring point and each child monitoring point; using the zero-order trend consistency deviation as the zero-order conservation residual; when the abnormal response chain corresponding to a certain electrical topology constraint candidate edge cannot reduce the zero-order conservation residual of the relevant parent monitoring point, reducing the credibility of the electrical topology constraint candidate edge or reducing the comprehensive score of the main anomaly source of the corresponding candidate monitoring point.

[0011] In one aspect of the present invention, the method for generating the operational transient occlusion result includes: detecting the duration of abnormal response, the time-series cumulative value, and the branch mismatch index of each monitoring point; when the duration of abnormal response of a certain monitoring point is less than the threshold for abnormal response duration, the time-series cumulative value is less than the threshold for time-series cumulative value, and the branch mismatch index is less than the threshold for branch mismatch index, the abnormal response of the monitoring point within the corresponding time window is marked as an operational transient response; the operational transient response is not used as a direct basis for determining the main anomaly source monitoring point, but forms an occlusion penalty when calculating the comprehensive score of the main anomaly source of the corresponding monitoring point.

[0012] In one aspect of the invention, calculating the comprehensive score of the main anomaly source for each monitoring point includes: calculating the net output score of directional information flow, the delay prior score, the electrical topology reliability score, the zero-order conservation residual penalty score, and the operational transient masking penalty score for each monitoring point; wherein, the net output score of directional information flow is jointly determined by the outward directional information flow of the monitoring point to other monitoring points and the inward directional information flow of other monitoring points to the monitoring point; the delay prior score is determined by the corrected response delay sequence of the monitoring point relative to other monitoring points; and the electrical topology reliability score is determined by the electrical topology associated with the monitoring point. The number of candidate edges for constraint is determined. The zero-order conservation residual penalty score is determined by the electrical consistency deviation of the anomaly response chain corresponding to the monitoring point. The operational transient masking penalty score is determined by the proportion of the monitoring point marked as an operational transient response. The normalized directional information flow net output score, delay pre-emptive score, electrical topology credibility score, zero-order conservation residual penalty score, and operational transient masking penalty score are weighted and fused to obtain the main anomaly source comprehensive score. The monitoring point with the largest main anomaly source comprehensive score is determined as the main anomaly source monitoring point.

[0013] In another aspect of the present invention, the present invention also relates to a leakage current fault tracing data processing system based on time-series correlation, for implementing the above-mentioned leakage current fault tracing data processing method based on time-series correlation, the system comprising a leakage current sensor network, a data acquisition unit, a time-series data processing engine, a time-series correlation analysis module, a main anomaly source identification engine, and a tracing report generation module; The leakage current sensor network is used to passively collect leakage current signals from multiple monitoring points under the condition that the power distribution network maintains its original operating state and no test signals are injected into the power distribution lines or the switches are controlled to open or close. The data acquisition unit is used to convert the leakage current signal into a digital leakage current sample value with a unified timestamp, and output the digital leakage current sample value to the time-series data processing engine. The time-series data processing engine is used to decouple the dual-branch leakage characteristics of the digital leakage current sampling value, generate trend characteristics, power frequency characteristics and branch mismatch indicators for each monitoring point, and output the trend characteristics, power frequency characteristics and branch mismatch indicators to the time-series correlation analysis module and the main anomaly source identification engine. The time-series correlation analysis module is used to calculate the directional information flow and corrected response delay between each monitoring point based on the trend characteristics, and output the directional information flow and the corrected response delay to the main anomaly source identification engine; The main anomaly source identification engine is used to calculate the comprehensive score of the main anomaly source for each monitoring point based on the directional information flow, the corrected response delay, electrical topology constraint relationship, parent-child loop zero-sequence conservation residual, branch mismatch index and operation transient masking result, and to determine the main anomaly source monitoring point and the anomaly response timing chain. The source tracing report generation module is used to generate a leakage fault source tracing report based on the main anomaly source monitoring point and the anomaly response timing chain.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention passively collects leakage current time-series data from multiple monitoring points while maintaining the original operating state of the power distribution network. It eliminates the need to inject test signals into the distribution lines or control switches to open or close, thus reducing the impact of fault tracing on the original power supply status. By establishing electrical topology constraints between multiple monitoring points, using parent-child circuit relationships, same-level branch relationships on the same busbar, and common grounding path relationships as conditions for generating candidate abnormal response edges, it reduces abnormal response relationships generated solely by statistical correlations that do not conform to actual electrical connections. By decoupling the leakage current time-series data into dual-branch leakage characteristics, trend characteristics, power frequency characteristics, and branch mismatch indices are obtained, distinguishing between persistent leakage trend changes and power frequency periodic disturbances, providing a clearer data foundation for subsequent main anomaly source identification. By calculating directional information flow and corrected response delay, and using measurement link response delay to correct observation response delay, the impact of differences in response from different sensors, signal conditioning circuits, and data acquisition channels on the determination of the sequence of abnormal responses can be reduced. By introducing the zero-sequence conservation residual of the parent-child loops and performing electrical consistency verification on candidate anomaly response chains, the scores of candidate monitoring points that are statistically correlated but cannot explain the leakage current trend relationship of the parent-child loops can be reduced. By using the results of operational transient shielding, abnormal responses caused by short-term load switching, switching operations, or non-faulty power frequency disturbances are deducted, reducing the likelihood of short-term operational disturbances being identified as the primary anomaly source. Through these processes, the system can output the primary anomaly source monitoring point and the anomaly response time sequence chain, providing data for maintenance personnel to determine the troubleshooting sequence and the scope of associated loops. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart of a leakage fault source tracing data processing method based on time-series correlation according to the present invention.

[0017] Figure 2 This is a flowchart of step 1 in the leakage fault source tracing data processing method based on time-series correlation of the present invention.

[0018] Figure 3 This is a flowchart of step 2 in the leakage fault source tracing data processing method based on time-series correlation of the present invention.

[0019] Figure 4This is a flowchart of step 3 in the leakage fault source tracing data processing method based on time-series correlation of the present invention.

[0020] Figure 5 This is a flowchart of step 4 in the leakage fault source tracing data processing method based on time-series correlation of the present invention.

[0021] Figure 6 This is a framework diagram of a leakage fault tracing data processing system based on time-series correlation according to the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to embodiments. These embodiments are merely some, not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the protection scope of the present invention.

[0023] Please see Figure 1 As shown, this embodiment discloses a method and system for processing leakage fault source tracing data based on time-series correlation, wherein the method specifically includes: S1. Under the condition that the power distribution network maintains its original operating state and no test signals are injected into the power distribution lines, and no switches are controlled to perform opening or closing, the leakage current timing data of multiple monitoring points are passively collected, and the electrical topology constraint relationship between multiple monitoring points is established. S2. Decouple the leakage current time series data of the two branches to obtain the trend characteristics, power frequency characteristics and branch mismatch index of each monitoring point. S3. Calculate the directional information flow and corrected response delay between each monitoring point based on trend characteristics, and generate candidate edges for electrical topology constraints in combination with electrical topology constraint relationships; S4. Calculate the comprehensive score of the main anomaly source for each monitoring point based on the candidate edges of electrical topology constraints, branch mismatch index, zero-sequence conservation residuals of parent and child loops, and operation transient occlusion results. Then, determine the main anomaly source monitoring point and anomaly response timing chain based on the comprehensive score of the main anomaly source.

[0024] In this embodiment, it should be noted that in S1, to address the issues that actively injecting test signals may alter the line's operating state, that switching on / off may cause power supply disturbances, and that single-point leakage current sampling is insufficient to reflect the sequential relationship of multiple circuit anomalies, the system passively collects leakage current time-series data from multiple monitoring points for processing. Specifically, the leakage current sensor network only samples the leakage current at each monitoring point, without outputting test signals to the line, controlling the operation of the distribution switch, or using state changes before and after switch operation as a basis for fault location. Simultaneously, the system establishes electrical topology constraints based on the actual power supply connection relationship of the distribution network, ensuring that subsequently generated candidate anomaly response edges must conform to parent-child loop relationships, same-level branch relationships under the same bus, or common grounding path relationships. This processing ensures that the subsequently calculated time-series correlation is no longer a statistical correlation between arbitrary monitoring points, but rather a candidate correlation constrained by electrical connection relationships.

[0025] In S2, to address the issue that the original leakage current signal simultaneously contains both power frequency periodic components and low-frequency trend components, a dual-branch leakage feature decoupling process is employed. Specifically, the trend feature branch performs effective value processing, smoothing, and trend extraction on the leakage current time-series data to obtain trend features reflecting a continuous increase or slow change in leakage current. The power frequency feature branch performs frequency domain processing on the same leakage current time-series data to obtain power frequency features reflecting changes in the fundamental and second harmonic frequencies. The system further generates a branch mismatch index based on the relative relationship between the trend features and the power frequency features. The branch mismatch index is used to determine whether an anomaly at a monitoring point is closer to a trend-based leakage current anomaly or closer to a normal power frequency disturbance or short-term operational fluctuation. Through this processing, the system avoids directly treating power frequency disturbances as leakage sources and also avoids using low-frequency data after trend filtering to extract power frequency features, which would lead to inconsistencies.

[0026] In S3, to address the issue of spurious causal edges that may arise from relying solely on directional information flow, candidate edges for electrical topology constraints are generated by combining directional information flow, corrected response delay, and electrical topology constraints. Specifically, the time-series correlation analysis module first calculates the directional information flow between monitoring points based on the trend characteristics of each monitoring point, and then obtains the observation response delay based on the response sequence of the trend characteristics. Since the leakage current sensors, signal conditioning circuits, and data acquisition channels at different monitoring points have different measurement link response delays, the system further corrects the observation response delay using pre-calibrated measurement link response delays to obtain the corrected response delay. Only when the directional information flow exceeds the directional information flow edge establishment threshold, the corrected response delay is positive, and there is an electrical topology constraint relationship between the two monitoring points, does the system retain the corresponding candidate edge as an electrical topology constraint candidate edge.

[0027] In S4, to address the issue that relying solely on directional information flow or graph sorting algorithms cannot fully reflect the electrical mechanism of leakage faults, a comprehensive scoring of the main anomaly source is used. Specifically, the main anomaly source identification engine calculates the outward anomaly impact capability of monitoring points based on candidate edges of electrical topology constraints, calculates the pre-emptiveness of monitoring points based on the corrected response delay, calculates the reliability of electrical topology based on the number of candidate edges of electrical topology constraints, calculates electrical consistency deviation based on the zero-order conservation residuals of parent and child loops, and deducts short-term operational disturbances based on the operational transient masking results. The main anomaly source identification engine combines the above factors into a comprehensive scoring of the main anomaly source, identifies the monitoring point with the highest score as the main anomaly source monitoring point, and generates an anomaly response time sequence chain from the candidate edges of electrical topology constraints, starting from the main anomaly source monitoring point. Therefore, the final output result is not a single alarm point, but a source tracing result including the main anomaly source monitoring point and the anomaly response time sequence chain.

[0028] like Figure 2 As shown, in one specific embodiment, S1 includes: S11, determining multiple monitoring points in the power distribution network; S12, passively collecting leakage current signals from each monitoring point through a leakage current sensor network; S13, generating digital leakage current sampling values ​​with a unified timestamp through a data acquisition unit, and forming leakage current time-series data; S14, establishing electrical topology constraint relationships based on the power supply connection relationships of the power distribution network.

[0029] In S11, monitoring points are set up based on the actual power supply hierarchy of the distribution network. Monitoring points are installed at the main incoming line, bus branch ends, branch power supply circuit outgoing ends, power supply ends of important loads, or locations with a high historical risk of leakage. Each monitoring point is assigned a unique number, and the corresponding distribution box number, circuit name, upstream circuit number, downstream circuit number, grounding method, and installation location are recorded. The monitoring point number is not used to represent the device communication address, but rather to represent the sampling location within the electrical circuit. This method allows the subsequently calculated abnormal response relationships to be mapped back to specific circuits, rather than remaining only at the network node or display node level.

[0030] In S12, the leakage current sensor network passively acquires leakage current signals from each monitoring point. The leakage current sensor can be a zero-sequence current transformer, installed outside the phase and neutral lines of the corresponding circuit. When there is no leakage anomaly in the corresponding circuit, the vector sum of the phase and neutral currents is close to zero; when there is a leakage anomaly in the corresponding circuit, part of the current flows to the grounding terminal or equipment casing through an abnormal path, and the zero-sequence current transformer outputs a secondary current signal related to the leakage current. This acquisition process does not apply external excitation to the power distribution line, does not change the load operating state, and does not control switch operation. The transformation ratio, accuracy class, and response time of the leakage current sensor are obtained through the factory test report or on-site calibration records and stored in the data storage module.

[0031] In S13, the data acquisition unit converts the signal output from the leakage current sensor into a digital leakage current sample value after signal conditioning. The signal conditioning process includes current-to-voltage conversion, amplitude conditioning, input protection, and necessary isolation processing. The data acquisition unit uses a multi-channel synchronous sampling analog-to-digital converter, enabling multiple monitoring points to obtain sample values ​​under the same sampling clock. Each sample value is appended with a uniform timestamp, forming leakage current time-series data. This leakage current time-series data is of the same type and does not include response data obtained from actively injected test signals, impedance spectrum data, or state transition data obtained from switch opening or closing disturbances. The sampling rate is determined by the hardware parameters of the data acquisition unit and the actual leakage response time requirements. For example, a sampling rate of 1000 times per second can be used in low-voltage distribution circuits. This sampling rate is derived from the sampling device configuration record and on-site debugging record.

[0032] In S14, the system establishes electrical topology constraints based on the power supply connections of the distribution network. These constraints include parent-child loop relationships, same-level branch relationships under the same busbar, and common grounding path relationships. Parent-child loop relationships are derived from the primary wiring diagram of the distribution system, the circuit list of the distribution box, and on-site wiring verification records. Same-level branch relationships under the same busbar are derived from the busbar connection records within the distribution box. Common grounding path relationships are derived from the grounding system design data, grounding terminal connection records, and on-site grounding resistance test records. The system generates electrical topology adjacency relationships based on the above data and uses them as constraints for subsequently generating candidate edges for electrical topology constraints. If no electrical connection relationship exists between two monitoring points, even if a strong temporal correlation is subsequently calculated between them, candidate edges for electrical topology constraints are not directly generated.

[0033] The shared grounding path relationship does not refer to the generalized shared grounding state where all circuits in the power distribution system are connected to the protective grounding system. Instead, it refers to a local connection relationship where corresponding circuits at two monitoring points share the same grounding busbar, the same grounding branch, the same grounding terminal block section, or a shared grounding path confirmed by on-site grounding connection records. If two circuits only have a remote shared grounding relationship at the building's main grounding point, but do not share a local grounding branch or the same grounding terminal block section, then this shared grounding path relationship is not recorded as an allowed edge construction relationship in the electrical topology adjacency matrix.

[0034] In some optional embodiments, a low-voltage distribution room includes one main incoming circuit, three busbar branch circuits, and eight outgoing circuits. The system sets monitoring points at 12 locations within this room. Monitoring point 1 corresponds to the main incoming circuit, monitoring points 2 to 4 correspond to the busbar branch circuits, and monitoring points 5 to 12 correspond to the outgoing circuits. According to the circuit list of the distribution box, monitoring point 1 has a parent-child circuit relationship with monitoring points 2 to 4, monitoring point 2 has a parent-child circuit relationship with monitoring points 5 to 7, and monitoring points 5 to 7 have a sibling branch relationship under the same busbar. The system writes the above relationships into the electrical topology constraint relationship table. Subsequently, when a statistical correlation occurs between monitoring point 5 and monitoring point 9, since they do not belong to the electrical connection relationship that allows edge creation, the system does not directly include them in the electrical topology constraint candidate edge.

[0035] The measurement link response delay is obtained based on a limited number of historical calibration data. Specifically, during on-site shutdown maintenance or factory testing, the same standard pulse signal is applied to the input terminal of each leakage current sensor. The time required for each channel to reach the set response ratio from the occurrence of the input signal is recorded. The average of multiple test results is used as the measurement link response delay for that monitoring point. The set response ratio can be 90% of the final stable value, derived from the sensor response time test specification or the company's internal sampling link calibration specification. The measurement link response delay of each monitoring point is stored in the data storage module for subsequent correction of the observed response delay.

[0036] It should be noted that the measurement link response delay calibration, controlled load switching test, and standard pulse test are only performed during factory testing, shutdown maintenance, or system commissioning and debugging, and are used to generate historical calibration parameters in the system parameter table. During the leakage fault tracing operation phase, the system only passively collects leakage current timing data from each monitoring point, does not inject test signals into the power distribution line, does not control the switch to open or close, and does not use artificially created switch state jumps as the basis for identifying the main abnormal source.

[0037] like Figure 3 As shown, in one specific embodiment, S2 includes: S21, performing trend feature branch processing on the leakage current time series data to obtain trend features; S22, performing power frequency feature branch processing on the leakage current time series data to obtain power frequency features; S23, generating branch mismatch index based on trend features and power frequency features; S24, combining trend features, power frequency features, and branch mismatch index into multi-dimensional time series feature data.

[0038] In S21, the trend feature branch is used to extract the component of leakage current that changes continuously over time. The system first performs sliding RMS processing on the leakage current time-series data of each monitoring point to reduce the influence of the instantaneous phase of the power frequency waveform on trend judgment. The window length of the sliding RMS processing is determined by the power frequency cycle. For example, when the power frequency is 50 Hz and the sampling rate is 1000 times per second, one power frequency cycle corresponds to 20 sampling points, so the sliding RMS window can be set to 20 sampling points. The source of this window length is the power frequency and sampling rate configuration of the power grid. After obtaining the sliding RMS, the system performs trend smoothing processing to obtain the trend sample value. The cutoff frequency of the trend smoothing processing can be set to 10 Hz. This cutoff frequency comes from the statistical results of leakage current trend change frequency band in historical leakage fault samples and is verified by on-site normal load fluctuation samples. The trend features include trend sample value, time-series change rate, and time-series cumulative value. The time-series change rate is used to reflect the rate of increase of leakage current, and the time-series cumulative value is used to reflect the persistence of leakage anomaly.

[0039] In S22, the power frequency characteristic branch is used to extract the periodic characteristics of the leakage current. The system does not directly apply a low cutoff frequency filter to the original leakage current time-series data before extracting the power frequency characteristics; instead, it performs frequency domain analysis in the branch that retains the power frequency components. The power frequency characteristics include at least the fundamental leakage current amplitude and the second harmonic leakage current amplitude. The fundamental leakage current amplitude reflects the main power frequency components of the leakage current, while the second harmonic leakage current amplitude reflects harmonic variations caused by nonlinear loads or rectifier-type loads. The fundamental frequency is 50 Hz, derived from the rated power frequency of the low-voltage distribution system. The second harmonic frequency is 100 Hz, derived from twice the fundamental frequency. If the project area uses a different power frequency, the fundamental frequency and second harmonic frequency are determined according to the local power grid's rated frequency and registered in the system parameters.

[0040] In S23, the branch mismatch index is used to characterize the prominence of trend anomalies relative to power frequency periodic disturbances. If the trend change rate of a monitoring point increases significantly, but the fundamental leakage current amplitude and the second harmonic leakage current amplitude do not show corresponding significant periodic disturbances, then the monitoring point is more likely to have a trend leakage anomaly. If a monitoring point mainly exhibits power frequency amplitude fluctuations, but the trend change rate and time-series cumulative value are low, then the monitoring point is more likely to be affected by ordinary load changes or short-term operational disturbances. The branch mismatch index threshold is derived from historical sample calibration. Specifically, the system collects at least two types of samples: one type is real leakage anomaly samples confirmed by manual inspection, and the other type is non-fault samples formed by switch operations, load switching, and normal power fluctuations. By comparing the distribution of branch mismatch indices in the two types of samples, the boundary value that can distinguish between the two types of samples is selected as the branch mismatch index threshold. If the number of samples is insufficient, the empirical threshold during the field commissioning phase is used first, and then updated after historical samples are accumulated.

[0041] In S24, the system aligns trend characteristics, power frequency characteristics, and branch mismatch indicators according to monitoring point numbers and timestamps to form multidimensional time-series characteristic data. This data serves as input for S3 to calculate directional information flow and corrected response delay. The multidimensional time-series characteristic data includes at least the trend sample value, time-series rate of change, time-series cumulative value, fundamental leakage current amplitude, second harmonic leakage current amplitude, and branch mismatch indicator for each monitoring point. The system retains the original timestamps when generating the multidimensional time-series characteristic data for subsequent determination of the response sequence between monitoring points. If a monitoring point has missing data at a certain timestamp, the system uses short-time interpolation of adjacent timestamps to fill the gap and records this completion status in the data quality marker. The data quality marker is not directly used as the basis for determining the main anomaly source, but it can be used to indicate the reliability of the sampling in the report.

[0042] In some optional embodiments, monitoring point 3 shows a continuous increase in its trend sampling value within a 10-second sliding window, with a time-series change rate exceeding the historical normal fluctuation range and a continuously increasing time-series cumulative value. Simultaneously, the fundamental leakage current amplitude at monitoring point 3 shows some change, but the second harmonic leakage current amplitude does not show a change of the same magnitude as the trend increase. The system generates a high branch mismatch index based on this trend characteristic and power frequency characteristics, and inputs this index into the subsequent main anomaly source identification engine. Monitoring point 5 shows a short-term power frequency amplitude change within the same window, but the trend sampling value has a short duration, a low time-series cumulative value, and a low branch mismatch index; the system will deduct this value during subsequent transient masking processing.

[0043] like Figure 4 As shown, in one specific embodiment, S3 includes: S31, calculating the directional information flow between each monitoring point based on trend characteristics; S32, calculating the observation response delay based on the response sequence relationship between trend characteristics; S33, correcting the observation response delay based on the measurement link response delay to obtain the corrected response delay; S34, generating electrical topology constraint candidate edges by combining the directional information flow, the corrected response delay, and the electrical topology constraint relationship.

[0044] In S31, the directional information flow is used to describe the directional interpretability of the trend characteristic change of one monitoring point on the subsequent trend characteristic change of another monitoring point. The system uses the trend sampling value output by S2 as the basis for calculation, rather than the original power frequency instantaneous waveform. The reason for this is that the phase difference of the original power frequency waveform may lead to errors in direction judgment, while the trend sampling value is more suitable for expressing the sequence of changes in leakage anomalies. The boundary threshold of the directional information flow is derived from the statistical distribution of historical normal operating condition samples and manually confirmed leakage samples. Specifically, the system first calculates the directional information flow distribution between each monitoring point in the historical normal operating condition samples, and uses the higher quantile value as the initial noise upper limit; then it verifies whether the threshold can retain known anomaly response edges in the manually confirmed leakage samples. If the threshold is too low, resulting in too many candidate edges under normal operating conditions, it is increased; if the threshold is too high, resulting in the truncation of known leakage anomaly chains, it is decreased.

[0045] In S32, the observation response delay is used to describe the temporal relationship between abnormal responses occurring at two monitoring points. The system determines the observation response delay of one monitoring point relative to the other based on the time-delay correlation results between the trend characteristic sequences of the two monitoring points. The observation response delay only represents the chronological relationship directly observed from the data and does not exclude the influence of response delays from different measurement links; therefore, it cannot be directly used as the basis for constructing candidate abnormal response edges. The search range for the observation response delay is derived from historical records of leakage current abnormal responses and the electrical distance of the distribution circuit. For example, for internal circuits in low-voltage distribution rooms, the search range can be set to tens to hundreds of milliseconds, derived from historical fault waveform records and sensor response time calibration records.

[0046] In S33, the system uses the measurement link response delay to correct the observation response delay, resulting in a corrected response delay. The measurement link response delay includes the leakage current sensor response delay, the signal conditioning circuit delay, and the data acquisition channel delay. These delays are obtained through factory testing, field calibration, or synchronization pulse testing. If the measurement link response delays of two monitoring points are different, directly using the observation response delay may lead to a reversal of the direction. By using the corrected response delay, the system can determine which monitoring point exhibits an abnormal response earlier after deducting the influence of the measurement link. Only when the corrected response delay is positive does the system consider the abnormal response direction from the earlier monitoring point to the later monitoring point to have the necessary timing conditions.

[0047] In step S34, the system combines directional information flow, corrected response delay, and electrical topology constraints to generate candidate edges for electrical topology constraints. Specifically, if the directional information flow between two monitoring points exceeds the directional information flow edge establishment threshold, the corrected response delay is positive, and there is an electrical connection marker allowing edge establishment between the two monitoring points in the electrical topology constraint relationship, then this directional edge is retained as a candidate edge for electrical topology constraints. If any of the three conditions is not met, no candidate edge for electrical topology constraints is generated. In this way, the system can delete pseudo-abnormal response edges that are only statistically correlated but do not satisfy the electrical connection relationship, and it can also delete erroneous directional edges caused by differences in measurement link delay.

[0048] In some optional embodiments, if the directional information flow between monitoring point 3 and monitoring point 4 exceeds the directional information flow edge establishment threshold, the corrected response delay is 120 milliseconds, and there is an allowed edge establishment relationship between monitoring point 3 and monitoring point 4 under the same bus, then the system retains the electrical topology constraint candidate edge from monitoring point 3 to monitoring point 4. Monitoring point 3 and monitoring point 9 may also exhibit a high statistical correlation within a certain window, but since there is no parent-child loop relationship, same-level branch relationship under the same bus, or common grounding path relationship between them, the system does not generate an electrical topology constraint candidate edge from monitoring point 3 to monitoring point 9.

[0049] like Figure 5 As shown, in one specific implementation, S4 includes: S41, calculating the zero-sequence conservation residual of the parent-child loop based on the parent-child loop relationship; S42, generating the operational transient masking result based on the abnormal response duration, time-series cumulative value, and branch mismatch index; S43, calculating the comprehensive score of the main anomaly source based on the directional information flow, electrical topology constraint candidate edges, corrected response delay, zero-sequence conservation residual of the parent-child loop, and operational transient masking result; S44, determining the main anomaly source monitoring point based on the comprehensive score of the main anomaly source, and generating the anomaly response time-series chain and source tracing report.

[0050] In S41, the zero-sequence conservation residual of the parent-child loop is used to determine whether a certain abnormal response relationship conforms to the electrical consistency between the parent and child loops. For monitoring points with a parent-child loop relationship, the trend change detected by the parent monitoring point should be explainable by one or more trend changes in its lower-level loops. If a candidate abnormal response chain can explain the trend change of the parent loop, the corresponding zero-sequence conservation residual is low; if a candidate abnormal response chain has statistical correlation but cannot explain the trend relationship between the parent and child loops, the corresponding zero-sequence conservation residual is high. The proportional correction coefficient comes from the historical fitting results of the parent-child loop trend data under normal operating conditions, or from the parent-child loop sampling ratio relationship obtained during field calibration. The fitted data should exclude known fault periods and sensor abnormal periods.

[0051] This implementation retains a computational relationship that must be disclosed, namely, the zero-order conservation residual of the parent-child loop is: ; in, Indicates the first The parent monitoring point at the first Zero-order conserved residuals of parent-child loops at each sampling time Indicates the first The parent monitoring point at the first Trend sampled values ​​at each sampling time, Indicates the first The set of child monitoring points corresponding to each parent monitoring point Indicates the sub-level monitoring point number. Indicates the first The parent monitoring point and the first The proportional correction coefficient between sub-level monitoring points Indicates the first Each sub-level monitoring point in the first Trend sampled values ​​at each sampling time, This represents a small constant used to prevent the denominator from being zero. This represents the absolute value operation. Indicates the first Sum the values ​​of all child monitoring points corresponding to each parent monitoring point. Indicates the parent monitoring point number. Indicates the sampling time number.

[0052] In S42, the operational transient masking results are used to identify short-term operational disturbances. The system detects the duration of abnormal responses, the time-series cumulative value, and the branch mismatch index for each monitoring point. If the duration of an abnormal response at a monitoring point is less than the abnormal response duration threshold, the time-series cumulative value is less than the time-series cumulative value threshold, and the branch mismatch index is less than the branch mismatch index threshold, then the abnormal response at that monitoring point within the corresponding time window is marked as an operational transient response. The abnormal response duration threshold is derived from the upper limit statistical value of the duration of non-fault transient responses in the field switch operation records and load switching records, with a preset margin added. The time-series cumulative value threshold is derived from the cumulative value distribution of historical normal operating conditions and non-fault operation samples. The branch mismatch index threshold is derived from the comparative calibration of actual leakage current abnormal samples and non-fault operation samples. The source, calibration time, and applicable circuit range of all the above thresholds should be recorded in the system parameter table.

[0053] In S43, the system calculates the comprehensive score of the main anomaly source for each monitoring point. The comprehensive score is jointly determined by the net output capability of directional information flow, the degree of delay precedence, the reliability of the electrical topology, the zero-order conservation residual penalty of the parent-child loop, and the operational transient masking penalty. The net output capability of directional information flow reflects the monitoring point's ability to interpret anomaly responses from other monitoring points. The degree of delay precedence reflects how early the monitoring point appears among multiple anomaly responses. The reliability of the electrical topology reflects whether the candidate edges associated with the monitoring point are supported by electrical topology constraints. The zero-order conservation residual penalty of the parent-child loop is used to deduct candidate sources that do not conform to the electrical consistency of the parent-child loop. The operational transient masking penalty is used to deduct candidate sources that are more likely to be caused by short-term operational disturbances.

[0054] This implementation retains another calculation relationship that must be disclosed, namely, the comprehensive score of the main anomaly source is: ; in, Indicates the first Comprehensive score of the main anomaly source at each monitoring point Indicates the first Net output score of directional information flow at each monitoring point Indicates the first Pre-calculation score for latency at each monitoring point Indicates the first Electrical topology reliability score for each monitoring point Indicates the first Zero-order conservation residual penalty score for parent-child loops at each monitoring point Indicates the first Operational transient occlusion penalty score for each monitoring point This indicates the weight of the net output score for directional information flow. This indicates the weight of the delay priority score. Indicates the weight of the electrical topology credibility score. This represents the weight of the zero-order conservation residual penalty score for the parent-child loop. This indicates the weight of the transient occlusion penalty score. This indicates the monitoring point number.

[0055] in, , , , and The weights are derived from the calibration results of historical leakage current samples and non-fault disturbance samples. Specifically, the system adjusts the weights in the historical samples so that the main anomaly source monitoring point in the known real leakage current anomaly samples scores higher than other monitoring points, while simultaneously reducing the scores of short-term anomaly points in the non-fault operation disturbance samples. The initial weights can be determined from the samples during the field commissioning phase, and can be updated subsequently based on new samples without changing the processing logic defined in the claims. Weight update records should be stored in the data storage module.

[0056] In S44, the system identifies the monitoring point with the highest comprehensive score as the primary anomaly source monitoring point. Starting from this primary anomaly source monitoring point, it selects candidate edges from the electrical topology constraint candidate edges that pass the parent-child loop zero-sequence conservation residual check and are not obscured by operational transients, forming an anomaly response timing chain. Each edge in the anomaly response timing chain records the starting monitoring point, ending monitoring point, directional information flow magnitude, corrected response delay, whether it satisfies electrical topology constraints, whether it passes the parent-child loop zero-sequence conservation residual check, and whether it is affected by operational transient obscuration. The source tracing report generation module generates a source tracing report based on the primary anomaly source monitoring point and the anomaly response timing chain.

[0057] In some optional embodiments, within a 10-second sliding window, the system obtains that the net output score of the directional information flow, the delay precession score, and the electrical topology credibility score of monitoring point 6 are all higher than those of other monitoring points within the same topological constraint range. Furthermore, when monitoring point 6 is considered a candidate primary anomaly source, the parent trend change of monitoring point 3 can be explained by the trend changes of monitoring point 6 and its sibling related monitoring points. The corresponding parent-child loop zero-order conservation residual is lower than the parent-child loop zero-order conservation residual threshold. Therefore, monitoring point 6 is not marked as an operational transient response within this time window. Although monitoring points 5 and 7 exhibit certain trend changes, their anomaly response duration is short, their time-series cumulative value is low, or they cannot reduce the zero-order conservation residual between monitoring point 3 and its child monitoring point set when considered as candidate primary anomaly sources. Therefore, their comprehensive primary anomaly source score is lower than that of monitoring point 6. The system ultimately identifies monitoring point 6 as the main anomaly source monitoring point and generates anomaly response timing chains from monitoring point 6 to monitoring point 3, from monitoring point 6 to monitoring point 5, and from monitoring point 6 to monitoring point 7. Each anomaly response edge satisfies the allowed edge construction mark in the electrical topology adjacency matrix, and the corresponding corrected response delay is positive.

[0058] In this implementation, all thresholds should have a clear source to avoid directly writing unverifiable parameters based on experience.

[0059] The directional information flow boundary threshold is derived from historical normal operating condition samples and manually confirmed leakage current samples. Specifically, the system statistically analyzes the directional information flow between monitoring points under historical normal operating conditions to obtain the background distribution under normal conditions; then, it statistically analyzes the abnormal response edges in the manually confirmed leakage current samples to obtain the directional information flow distribution under abnormal conditions. The directional information flow boundary threshold is set to a value that can distinguish between the normal background distribution and the abnormal response distribution, and is verified through on-site commissioning samples. If the circuit types within the same distribution room differ significantly, the directional information flow boundary threshold can be set separately for each circuit type.

[0060] The threshold for the duration of abnormal responses is derived from field switch operation records, load switching records, and non-fault short-term disturbance records. Specifically, the system calculates the upper limit of the duration of non-fault transient responses in the above records and sets a margin based on sampling error and measurement link response delay. This threshold is used to distinguish between short-term operational disturbances and persistent leakage anomalies. If a field lacks sufficient historical records, a limited number of controlled load switching tests can be performed during the commissioning phase, with the test results used as the source of the initial threshold.

[0061] The time-series cumulative value threshold is derived from the distribution of time-series cumulative values ​​of normal operating condition samples, non-fault operation samples, and confirmed leakage samples. Specifically, the system statistically analyzes the range of time-series cumulative values ​​under normal operation, load switching, and actual leakage conditions, and selects the value that excludes most non-fault operation samples while retaining actual leakage samples as the threshold. This threshold is used to determine whether the anomaly has a continuous cumulative effect.

[0062] The branch mismatch index threshold is derived from the joint calibration of trend characteristics and power frequency characteristics. Specifically, the system collects real leakage current samples and non-fault periodic disturbance samples, compares the distribution of branch mismatch indices of the two, and selects the boundary value that can distinguish between trend leakage current anomalies and periodic disturbances as the threshold. This threshold is used to judge transient blocking during operation and also to help explain whether a certain abnormal response is closer to a leakage current trend anomaly.

[0063] The zero-sequence conservation residual threshold for the parent-child loop is derived from the residual statistical results of historical normal samples and confirmed leakage samples of the parent-child loop. Specifically, under normal operating conditions, the system fits a proportional correction coefficient based on the trend sampling values ​​of the parent and child monitoring points and calculates the residual background range; in the confirmed leakage samples, the system statistically analyzes the residual changes corresponding to the actual abnormal response chains. This threshold is used to determine whether the candidate abnormal response chains meet electrical consistency requirements.

[0064] The early warning threshold is derived from normal background leakage current levels, historical alarm records, and on-site operation and maintenance rules. The early warning threshold can be compensated for based on background leakage current trend drift, but it must not alter the mandatory safety protection threshold corresponding to a Level 1 severe leakage current fault. The mandatory safety protection threshold is derived from applicable electrical safety regulations, leakage current protection device parameters, and operation and maintenance management procedures. The system records the early warning threshold and the mandatory safety protection threshold separately in the parameter table to avoid mixing the two types of thresholds.

[0065] The measurement link response delay originates from the calibration data of the sensor, signal conditioning circuit, and acquisition channels. Specifically, during factory testing or field calibration, the same standard pulse is applied to each sampling channel, and the time it takes for each channel to reach the set response ratio is recorded. The average of multiple tests is used as the measurement link response delay for that channel. The set response ratio is derived from the measurement link calibration specification, and a commonly used ratio is 90% of the final stable value. If the leakage current sensor, signal conditioning circuit, or data acquisition unit is replaced, the measurement link response delay should be recalibrated.

[0066] like Figure 6 As shown, in some embodiments, this embodiment discloses a leakage current fault tracing data processing system based on time-series correlation to implement the above-described method. The system includes a leakage current sensor network, a data acquisition unit, a time-series data processing engine, a time-series correlation analysis module, a main anomaly source identification engine, and a tracing report generation module. The system may also include a data storage module, a human-computer interaction module, an adaptive learning module, edge computing nodes, and a cloud server.

[0067] A leakage current sensor network is used to passively collect leakage current signals from multiple monitoring points while maintaining the original operating state of the power distribution network. The leakage current sensor network does not have the function of injecting test signals into the power distribution lines, nor does it output commands to control the opening and closing of switches. Its output is the leakage current signal of each monitoring point.

[0068] The data acquisition unit converts the leakage current signal into digital leakage current sample values ​​with a uniform timestamp. The data acquisition unit includes signal conditioning circuitry and a multi-channel synchronous sampling analog-to-digital converter. The digital leakage current sample values ​​output by the data acquisition unit serve as input to the timing data processing engine.

[0069] The time-series data processing engine is used to decouple the dual-branch leakage characteristics of the digitized leakage current sampling values. The engine outputs trend features, power frequency features, and branch mismatch indices. The trend features are output to the time-series correlation analysis module for calculating directional information flow and corrected response delay. The power frequency features and branch mismatch indices are output to the main anomaly source identification engine for operational transient masking and comprehensive scoring of the main anomaly source.

[0070] The time-series correlation analysis module is used to calculate the directional information flow and corrected response delay between monitoring points based on trend characteristics. The output of this module is not the final location result, but rather the input to the main anomaly source identification engine. This module also calls the measurement link response delay parameters from the data storage module to correct the observation response delay.

[0071] The main anomaly source identification engine calculates the comprehensive score of the main anomaly source for each monitoring point based on directional information flow, corrected response delay, electrical topology constraints, parent-child loop zero-sequence conservation residuals, branch mismatch indices, and operational transient masking results. The engine includes an electrical topology constraint candidate edge generation unit, a parent-child loop zero-sequence conservation residual verification unit, an operational transient masking unit, and a main anomaly source comprehensive scoring unit. The electrical topology constraint candidate edge generation unit removes pseudo-anomaly response edges that do not satisfy electrical connection relationships. The parent-child loop zero-sequence conservation residual verification unit determines whether candidate anomaly response chains satisfy parent-child loop electrical consistency. The operational transient masking unit identifies short-term non-fault disturbances. The main anomaly source comprehensive scoring unit determines the main anomaly source monitoring point and the anomaly response timing chain.

[0072] The source tracing report generation module is used to generate a leakage fault source tracing report based on the main anomaly source monitoring point and the anomaly response timing chain. The report includes the main anomaly source monitoring point number, corresponding circuit name, installation location, anomaly response timing chain, directional information flow results, corrected response delay, zero-sequence conservation residuals of parent and child circuits, operational transient shielding results, comprehensive score of the main anomaly source, and handling recommendations. The anomaly response timing chain in the report is an anomaly response chain under electrical topology constraints, not a communication path or an IoT topology path.

[0073] The data storage module is used to store raw leakage current sampling data, digital leakage current sampling values, trend characteristics, power frequency characteristics, branch mismatch indicators, electrical topology constraint relationships, measurement link response delay parameters, threshold source records, directional information flow results, corrected response delay, parent-child loop zero-sequence conservation residuals, operational transient shielding results, main anomaly source comprehensive score, and source tracing report.

[0074] The human-computer interaction module is used to display the electrical topology of the power distribution network, real-time leakage current at monitoring points, main anomaly source monitoring points, anomaly response timing chains, and historical source tracing reports. The module displays results under electrical topology constraints and does not use communication addresses and protocols as the basis for anomaly path generation.

[0075] The adaptive learning module is used to compensate for background drift in the early warning threshold based on historical normal operating condition data. The adaptive learning module does not change the mandatory safety protection threshold. The compensation parameters output by the adaptive learning module should be written to the data storage module, and adjustment records should be retained in the report.

[0076] Edge computing nodes can be deployed with time-series data processing engines, time-series correlation analysis modules, and main anomaly source identification engines. Cloud servers can be deployed with data storage modules, human-computer interaction modules, and source tracing report generation modules. Even if communication between the edge computing node and the cloud server is interrupted, the edge computing node can still identify the main anomaly source monitoring point based on local sampling data and local parameters, and upload the source tracing report after communication is restored.

[0077] In some optional embodiments, a power distribution network is equipped with 8 monitoring points. Monitoring point 1 is located at the main incoming line, monitoring points 2 to 4 are located at the outgoing lines of the sub-distribution boxes, and monitoring points 5 to 8 are located at the outgoing lines of the end load circuits. The system establishes electrical topology constraints based on the distribution box circuit list. Monitoring point 1 has a parent-child circuit relationship with monitoring points 2 to 4, monitoring point 3 has a parent-child circuit relationship with monitoring points 5 to 8, and monitoring points 5 to 8 have a sibling branch relationship on the same busbar.

[0078] At a certain moment, a leakage current anomaly occurred in the circuit corresponding to monitoring point 3. The leakage current sensor network passively collected leakage current signals from each monitoring point, and the data acquisition unit generated digital leakage current sample values ​​with a unified timestamp. After the time-series data processing engine processed the data of monitoring point 3 through the trend feature branch, it found that its trend sample value continued to rise, and both the time-series change rate and the time-series cumulative value were higher than the normal background range. The power frequency feature branch showed that there was a change in the amplitude of the fundamental leakage current at monitoring point 3, but this change was insufficient to explain the trend increase, so the branch mismatch index was high.

[0079] The time-series correlation analysis module calculates the directional information flow and corrected response delay based on trend characteristics. Monitoring point 3 exhibits directional information flow relative to monitoring points 4, 5, and 6, and its corrected response delay is positive. The main anomaly source identification engine further queries electrical topology constraints, confirming that monitoring point 3 has an electrical connection with the aforementioned monitoring points that allows edge establishment, thus generating candidate electrical topology constraints. Meanwhile, although there is a certain statistical correlation between monitoring points 2 and 7, they do not satisfy the electrical connection requirement for edge establishment, and the system deletes this candidate relationship.

[0080] The primary anomaly source identification engine calculates the zero-order conservation residuals of the parent-child loops. When monitoring point 3 is selected as the candidate primary anomaly source, its anomaly response chain can explain the trend changes in the upper and lower loops, and the zero-order conservation residuals of the parent-child loops are low. When monitoring point 4 is selected as the candidate primary anomaly source, although monitoring point 4 also exhibits some trend changes, its anomaly response chain cannot explain the trend changes in the lower loops of monitoring point 3, and the zero-order conservation residuals of the parent-child loops are high. Based on this, the system lowers the comprehensive score of monitoring point 4 as the primary anomaly source.

[0081] Within the same time window, monitoring point 6 experienced a short-term fluctuation, the duration of which was less than the anomaly response duration threshold, the time-series cumulative value was lower than the time-series cumulative value threshold, and the branch mismatch index was lower than the branch mismatch index threshold. The operational transient masking unit marked this fluctuation as an operational transient response and included a masking penalty in the main anomaly source comprehensive score of monitoring point 6. Ultimately, monitoring point 3 had the highest main anomaly source comprehensive score, and the system determined monitoring point 3 as the main anomaly source monitoring point, generating an anomaly response time-series chain starting from monitoring point 3.

[0082] The source tracing report generation module generates a report. The report records that the main anomaly source monitoring point is monitoring point 3, the corresponding circuit is the third branch power supply circuit, and the anomaly response timing chain includes monitoring points 3 to 4, 3 to 5, and 3 to 6. The report also lists the directional information flow results, corrected response delay, whether electrical topology constraints are met, zero-sequence conservation residuals of parent and child circuits, and operational transient shielding results for each anomaly response edge. Maintenance personnel should prioritize checking the insulation status and terminal block status of the circuit corresponding to monitoring point 3 based on the report.

[0083] To better enable those skilled in the art to fully understand and implement the technical solutions described in this specification, the following supplements the specific implementation principles of the technical solutions described in this specification with a specific application scenario. Assume that a low-voltage power distribution system in an office building includes one main distribution cabinet, three floor distribution boxes, and several terminal lighting and socket circuits. The main incoming circuit of the main distribution cabinet is designated as monitoring point 1; the outgoing circuit of the first-floor distribution box is designated as monitoring point 2; the outgoing circuit of the second-floor distribution box is designated as monitoring point 3; the outgoing circuit of the third-floor distribution box is designated as monitoring point 4; and the lighting circuit, socket circuit, and air conditioning terminal circuit on the second floor are designated as monitoring points 5, 6, and 7, respectively. Each monitoring point is equipped with a leakage current sensor. The output signal of the leakage current sensor is conditioned before entering the data acquisition unit. During system operation, no test signals are injected into the power distribution lines, and no circuit breakers or smart switches are controlled to perform opening or closing actions. Leakage current timing data of each monitoring point is collected only under the condition that the power distribution system maintains normal power supply.

[0084] Before the system was put into use, maintenance personnel established electrical topology constraints in the data storage module based on the primary wiring data, circuit number table, on-site cable routing, and grounding connection records of the office building's distribution boxes. Specifically, monitoring point 1 was recorded as a parent-child circuit with monitoring points 2, 3, and 4; monitoring point 3 was recorded as a parent-child circuit with monitoring points 5, 6, and 7; and monitoring points 5, 6, and 7 were recorded as sibling branches under the same floor's distribution box. For monitoring points without power supply connections, common busbar connections, or common grounding paths, the system did not allow them to directly generate electrical topology constraint candidate edges. The measurement link response delay of each leakage current sensor, signal conditioning circuit, and acquisition channel was obtained through standard pulse testing during the commissioning phase. Each channel was tested continuously for no less than 5 times, and the average value of the multiple test results was taken as the measurement link response delay of that monitoring point. This parameter was then bound to the monitoring point number and stored.

[0085] When a load branch in the second-floor socket circuit experiences leakage due to insulation aging, the leakage current time-series data at monitoring point 6 will initially show a continuous increase. Since monitoring point 6 is a downstream circuit of monitoring point 3, the leakage trend change may also be detected in the second-floor distribution box outgoing circuit corresponding to monitoring point 3. Simultaneously, due to the shared busbar and grounding path within the same floor distribution box, monitoring points 5 and 7 may also show smaller or delayed response changes. In this case, if only the leakage current at each monitoring point exceeds the warning threshold, the system may simultaneously obtain results indicating anomalies at monitoring points 3, 5, 6, and 7, but cannot directly determine which monitoring point corresponds to the primary anomaly source. The solution described in this manual does not directly use the monitoring point with the largest current value as the primary anomaly source. Instead, the time-series data processing engine first decouples the leakage current time-series data of each monitoring point using dual-branch leakage characteristics.

[0086] In the trend characteristic branch, the system performs effective value processing and trend processing on the leakage current time-series data of each monitoring point to obtain trend characteristics, time-series change rate, and time-series cumulative value. Monitoring point 6, due to continuous leakage in its corresponding circuit, shows a continuously rising trend characteristic, with the time-series cumulative value increasing over time. Monitoring points 5 and 7, if only affected by a common grounding path within the same distribution box or load disturbances, may show weaker trend changes and shorter durations. In the power frequency characteristic branch, the system extracts the fundamental leakage current amplitude and second harmonic leakage current amplitude of each monitoring point. If the anomaly at a certain monitoring point mainly originates from load switching or nonlinear load disturbances, its power frequency characteristics may change, but the trend characteristics and time-series cumulative value do not continuously increase. Based on the trend characteristics and power frequency characteristics, the system generates a branch mismatch index to distinguish between continuous leakage current trend anomalies and short-term power frequency disturbances.

[0087] Subsequently, the time-series correlation analysis module calculates the directional information flow and observation response delay based on the trend characteristics of each monitoring point, and corrects the observation response delay using pre-calibrated measurement link response delays. For example, if the trend change of monitoring point 6 is earlier than that of monitoring points 3, 5, and 7, and after deducting the response delays of each measurement channel, the corrected response delays from monitoring point 6 to monitoring point 3, from monitoring point 6 to monitoring point 5, and from monitoring point 6 to monitoring point 7 are still positive, then the above directional relationship meets the time-series conditions. The system further queries the electrical topology constraint relationships and confirms that there is a parent-child loop relationship between monitoring point 6 and monitoring point 3, and a sibling branch relationship between monitoring point 6 and monitoring points 5 and 7 under the same floor distribution box. Therefore, the above directional relationships can be retained as candidate edges for electrical topology constraints. If the system simultaneously finds that monitoring point 6 has a statistical correlation with a monitoring point located in the distribution box on the 1st floor, but there is no electrical connection between them that allows the establishment of an edge, then this statistical correlation relationship will not be retained as a candidate edge for electrical topology constraints.

[0088] During the primary anomaly source identification phase, the primary anomaly source identification engine further calculates the zero-order conservation residuals of the parent-child loops. For monitoring point 3 and its subordinate monitoring points 5, 6, and 7, the system compares the trend sampled value of monitoring point 3 with the proportionally corrected combination of the trend sampled values ​​of its subordinate monitoring points. If monitoring point 6 is selected as a candidate primary anomaly source, and its anomaly change can explain part of the response changes of monitoring point 3 and its peer branch monitoring points, then the corresponding zero-order conservation residuals of the parent-child loops are low. If monitoring point 5 is selected as a candidate primary anomaly source, but the trend change of monitoring point 5 has a short duration, low cumulative value, and cannot explain the main trend change of monitoring point 3, then the corresponding zero-order conservation residuals of the parent-child loops are high. Based on this, the system lowers the score of monitoring point 5 as the primary anomaly source.

[0089] Simultaneously, the system generates operational transient shielding results based on the duration of the abnormal response, the time-series cumulative value, and the branch mismatch index. For example, before and after a leakage current anomaly occurs, the air conditioning terminal circuit corresponding to monitoring point 7 may experience a short-term increase in leakage current due to compressor startup. However, if the duration of this increase is shorter than the abnormal response duration threshold, the time-series cumulative value is lower than the time-series cumulative value threshold, and the branch mismatch index is lower than the branch mismatch index threshold, then the system marks this abnormal response as an operational transient response. The aforementioned thresholds are derived from the calibration results of the office building's historical load switching records, normal operation samples, and manually confirmed leakage current samples, and the calibration time, sample quantity, and applicable circuit range are recorded in the system parameter table. Monitoring points marked as operational transient responses are not directly excluded from all calculations, but will incur shielding penalties in the main anomaly source comprehensive score.

[0090] Finally, the main anomaly source identification engine calculates the comprehensive score of the main anomaly source for each monitoring point by considering the net output of directional information flow, the sequence of corrected response delays, the reliability of the electrical topology, the zero-sequence conservation residuals of parent and child circuits, and the operational transient masking results. If the net output of directional information flow at monitoring point 6 is high, the early characteristics of the corrected response delay are clear, the relevant candidate edges satisfy the electrical topology constraints, and the zero-sequence conservation residuals of the parent and child circuits are low, and it is not marked as an operational transient response, then the system identifies monitoring point 6 as the main anomaly source monitoring point. Based on this, the source tracing report generation module generates a leakage fault source tracing report, which records the main anomaly source monitoring point as monitoring point 6, the corresponding circuit as the 2nd-level socket circuit, and lists the anomaly response sequence chain from monitoring point 6 to monitoring points 3, 5, and 7. Maintenance personnel can use this report to prioritize checking the insulation status, terminals, and end loads of the 2nd-level socket circuit, rather than checking all monitoring points with abnormal responses one by one. This specific application process demonstrates that the technical solution described in this manual can continuously combine passively collected leakage current time-series data from multiple monitoring points, electrical topology constraints, measurement link response delay, parent-child loop zero-sequence conservation residuals, and operational transient shielding results without altering the operating state of the power distribution network. This combination is used to determine the main anomaly source monitoring point and the anomaly response time-series chain.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A data processing method for tracing leakage current faults based on time-series correlation, characterized in that, include: Under the condition that the power distribution network maintains its original operating state and no test signals are injected into the power distribution lines, and no switches are controlled to perform opening or closing, leakage current time sequence data of multiple monitoring points are passively collected, and electrical topology constraint relationships between the multiple monitoring points are established. The leakage current time series data is decoupled from the dual-branch leakage characteristics to obtain the trend characteristics, power frequency characteristics and branch mismatch index of each monitoring point; Based on the trend characteristics, the directional information flow and corrected response delay between each monitoring point are calculated, and candidate edges for electrical topology constraints are generated in combination with the electrical topology constraint relationships. Based on the candidate edges of the electrical topology constraints, the branch mismatch index, the zero-sequence conservation residuals of the parent-child loops, and the operational transient occlusion results, the comprehensive score of the main anomaly source for each monitoring point is calculated, and the main anomaly source monitoring point and the anomaly response timing chain are determined according to the comprehensive score of the main anomaly source.

2. The leakage fault source tracing data processing method based on time-series correlation according to claim 1, characterized in that, Establishing the electrical topology constraint relationship between the multiple monitoring points includes: determining the parent-child loop relationship, the same-level branch relationship under the same bus, and the common grounding path relationship between each monitoring point according to the power supply connection relationship of the power distribution network; and generating an electrical topology adjacency matrix based on the parent-child loop relationship, the same-level branch relationship under the same bus, and the common grounding path relationship; wherein, generating an electrical topology constraint candidate edge between two monitoring points is only allowed when there is an electrical connection mark that allows edge establishment between two monitoring points in the electrical topology adjacency matrix.

3. The leakage fault source tracing data processing method based on time-series correlation according to claim 1, characterized in that, Passively acquiring leakage current timing data from multiple monitoring points includes: acquiring zero-sequence leakage current signals of the corresponding circuits through leakage current sensors deployed at each monitoring point; inputting the zero-sequence leakage current signals into a multi-channel synchronous sampling analog-to-digital converter after signal conditioning to obtain digital leakage current sampling values ​​with a unified timestamp; and forming a similar leakage current timing group by combining the digital leakage current sampling values ​​obtained from each monitoring point under the same sampling clock. The similar leakage current timing group does not include impedance spectrum data obtained from actively injected test signals, nor does it include state transition data obtained from switch opening or closing disturbances.

4. The leakage fault source tracing data processing method based on time-series correlation according to claim 1, characterized in that, The decoupling of leakage current characteristics in the time series data includes: in the trend characteristic branch, calculating the sliding effective value and trend filtering of the leakage current time series data to obtain trend sampling values, and calculating the time series change rate and time series cumulative value based on the trend sampling values; in the power frequency characteristic branch, performing spectrum analysis on the leakage current time series data to extract the fundamental leakage current amplitude and the second harmonic leakage current amplitude; and generating a branch mismatch index based on the time series change rate, the fundamental leakage current amplitude, and the second harmonic leakage current amplitude, which is used to characterize the prominence of trend anomalies relative to power frequency periodic disturbances.

5. The leakage fault source tracing data processing method based on time-series correlation according to claim 1, characterized in that, Calculating the directional information flow and corrected response delay between monitoring points based on the trend characteristics includes: constructing a state space reconstruction vector based on the trend sampling values ​​of each monitoring point, and calculating the normalized value of the transfer entropy between any two monitoring points based on the state space reconstruction vector, using the normalized value of the transfer entropy as the directional information flow value; determining the observation response delay based on the time delay correlation results between the trend sampling values ​​of the two monitoring points; correcting the observation response delay using the measurement link response delay of the leakage current sensor, signal conditioning circuit, and data acquisition channel corresponding to each of the two monitoring points, to obtain the corrected response delay; wherein, only when the directional information flow value exceeds the directional information flow edge establishment threshold and the corrected response delay is positive, the corresponding two monitoring points have the timing conditions to generate candidate abnormal response edges.

6. The leakage fault source tracing data processing method based on time-series correlation according to claim 5, characterized in that, The process of generating candidate edges for electrical topology constraints based on the electrical topology constraints includes: matching candidate abnormal response edges that satisfy the timing conditions with the electrical topology adjacency matrix; if there is an electrical connection marker between the starting and ending monitoring points of the candidate abnormal response edge that allows edge establishment, then the candidate abnormal response edge is retained as a candidate edge for electrical topology constraints; if there is no electrical connection marker that allows edge establishment, then the candidate abnormal response edge is deleted; thereby filtering out pseudo-abnormal response edges that only have statistical correlation but do not satisfy the electrical connection relationship of the distribution network.

7. The leakage fault source tracing data processing method based on time-series correlation according to claim 1, characterized in that, The generation method of the zero-order conservation residual of the parent-child loop includes: for monitoring points with parent-child loop relationship, calculating the zero-order trend consistency deviation between the parent monitoring point and the child monitoring point set based on the trend sampling value of the parent monitoring point, the trend sampling value of each child monitoring point in the child monitoring point set, and the proportional correction coefficient between the parent monitoring point and each child monitoring point; using the zero-order trend consistency deviation as the zero-order conservation residual; when the abnormal response chain corresponding to a certain electrical topology constraint candidate edge cannot reduce the zero-order conservation residual of the relevant parent monitoring point, reducing the credibility of the electrical topology constraint candidate edge or reducing the comprehensive score of the main anomaly source of the corresponding candidate monitoring point.

8. The leakage fault source tracing data processing method based on time-series correlation according to claim 1, characterized in that, The method for generating the operational transient occlusion result includes: detecting the duration of abnormal responses, the time-series cumulative value, and the branch mismatch index of each monitoring point; when the duration of abnormal responses of a certain monitoring point is less than the threshold for abnormal response duration, the time-series cumulative value is less than the threshold for time-series cumulative value, and the branch mismatch index is less than the threshold for branch mismatch index, the abnormal response of the monitoring point within the corresponding time window is marked as an operational transient response; the operational transient response is not used as a direct basis for determining the main anomaly source monitoring point, but forms an occlusion penalty when calculating the comprehensive score of the main anomaly source of the corresponding monitoring point.

9. The leakage fault source tracing data processing method based on time-series correlation according to claim 1, characterized in that, The calculation of the comprehensive score for the main anomaly source at each monitoring point includes: calculating the net output score of directional information flow, the delay prior score, the electrical topology reliability score, the zero-order conservation residual penalty score, and the operational transient masking penalty score for each monitoring point; wherein, the net output score of directional information flow is jointly determined by the outward directional information flow of the monitoring point to other monitoring points and the inward directional information flow of other monitoring points to the monitoring point; the delay prior score is determined by the order of the corrected response delays of the monitoring point relative to other monitoring points; and the electrical topology reliability score is determined by the electrical topology constraint candidates associated with the monitoring point. The number of edges is determined, the zero-order conservation residual penalty score is determined by the electrical consistency deviation of the anomaly response chain corresponding to the monitoring point, and the operational transient masking penalty score is determined by the proportion of the monitoring point marked as an operational transient response. The normalized directional information flow net output score, delay pre-emptive score, electrical topology reliability score, zero-order conservation residual penalty score, and operational transient masking penalty score are weighted and fused to obtain the main anomaly source comprehensive score. The monitoring point with the largest main anomaly source comprehensive score is determined as the main anomaly source monitoring point.

10. A leakage current fault tracing data processing system based on time-series correlation, characterized in that, The system is used to implement the leakage current fault tracing data processing method based on time-series correlation as described in any one of claims 1 to 9. The system includes a leakage current sensor network, a data acquisition unit, a time-series data processing engine, a time-series correlation analysis module, a main anomaly source identification engine, and a tracing report generation module. The leakage current sensor network is used to passively collect leakage current signals from multiple monitoring points under the condition that the power distribution network maintains its original operating state and no test signals are injected into the power distribution lines or the switches are controlled to open or close. The data acquisition unit is used to convert the leakage current signal into a digital leakage current sample value with a unified timestamp, and output the digital leakage current sample value to the time-series data processing engine. The time-series data processing engine is used to decouple the dual-branch leakage characteristics of the digital leakage current sampling value, generate trend characteristics, power frequency characteristics and branch mismatch indicators for each monitoring point, and output the trend characteristics, power frequency characteristics and branch mismatch indicators to the time-series correlation analysis module and the main anomaly source identification engine. The time-series correlation analysis module is used to calculate the directional information flow and corrected response delay between each monitoring point based on the trend characteristics, and output the directional information flow and the corrected response delay to the main anomaly source identification engine; The main anomaly source identification engine is used to calculate the comprehensive score of the main anomaly source for each monitoring point based on the directional information flow, the corrected response delay, electrical topology constraint relationship, parent-child loop zero-sequence conservation residual, branch mismatch index and operation transient masking result, and to determine the main anomaly source monitoring point and the anomaly response timing chain. The source tracing report generation module is used to generate a leakage fault source tracing report based on the main anomaly source monitoring point and the anomaly response timing chain.